An image sensor has a large number of sensor elements (pixels), generally greater than 1 million, in a Cartesian (square) grid. Conventional color image sensors are fabricated with colored filters arranged in a Bayer configuration. An Example of a convention Bayer configuration is illustrated in FIG. 6. The color scheme includes red, green, and blue filters (RGB). The Bayer filter pattern is 50% green, 25% red and 25% blue, hence is also referred to GRGB or other permutation such as RGGB. Twice as many green elements as red or blue are used to mimic the human eye's greater resolving power with green light. Since each pixel is filtered to record only one of three colors, the data from each pixel cannot fully determine color on its own. To obtain a full-color image, various demosaicing algorithms can be used to interpolate a set of complete red, green, and blue values for each point in the sensed scene.
Indeed, to obtain the full color image of the scene, data from all three color filters is required. Because data from all three color filters is required and each row of filters only has two types of color filter, at least two rows of pixels must be used to reproduce the scene using a Bayer configuration. This, in turn, has implications on the performance of image processing. Conventional digital image processors process one row at a time. Therefore, at least one row of sensor data must be held in a buffer while data from the next row is processed. In this manner, red, green, and blue data for each point in the image can be processed, however, it comes at the cost of processing speed.
One challenge of a designer of color image sensors is to consistently align the color response of the sensor pixels to the response curve of the human eye. Filter-based color sensors are device dependent. That is, different devices detect or reproduce different RGB values. The RGB values typically vary from manufacturer to manufacturer based on the manufacturer's selection of dye or phosphor used to make their filters. Further, filter degradation over time may even lead to variations in the RGB values over time in the same device.
The layers of a typical sensor are listed in Table I and shown in FIG. 1.
TABLE ITypical Layer Description Thickness (μm)15OVERCOAT2.0014MICRO LENS0.77313SPACER1.4012COLOR FILTER1.2011PLANARIZATION1.4010PASS30.6009PASS20.1508PASS11.007IMD5B0.3506METAL331.185IMD2B0.2004METAL221.183IMD1B0.2002METAL11.181ILD0.750
In Table I, typically the first layer on a silicon substrate is the ILD layer and the topmost layer is the overcoat. In Table I, ILD refers to a inter-level dielectric layer, METAL1, METAL2 and METAL3 refer to different metal layers, IMD1B, IMD2B and IMD5B refer to different inter-metal dielectric layers which are spacer layers, PASS1, PASS2 and PASS3 refer to different passivation layers (typically dielectric layers).
The total thickness of the layers above the silicon substrate of the image sensor is the stack height (s) of the image sensor and is the sum of the thickness of the individual layers. In the example of Table I, the sum of the thickness of the individual layers is typically about 11.6 micrometers (μm).
The space above the photosensitive element of a pixel must be transparent to light to allow incident light from a full color scene to impinge on the photosensitive element located in the silicon substrate. Consequently, no metal layers are routed across the photosensitive element of a pixel, leaving the layers directly above the photosensitive element clear.
The pixel pitch to stack height ratio (p/s) determines the cone of light (F number) that can be accepted by the pixel and conveyed to the photosensitive element on the silicon. As pixels become smaller and the stack height increases, this number decreases, thereby lowering the efficiency of the pixel.
More importantly, an increased stack height with greater number of metal layers obscure the light from being transmitted through the stack to reach the photosensitive element, in particular of the rays that impinge the sensor element at an angle. One solution is to decrease the stack height by a significant amount (i.e., >2 μm). However, this solution is difficult to achieve in a standard CMOS process.
Another issue, which possibly is the one that most limits the performance of the conventional image sensors, is that less than about one-third of the light impinging on the image sensor is transmitted to the photosensitive element such as a photodiode. In the conventional image sensors, in order to distinguish the three components of light so that the colors from a full color scene can be reproduced, two of the components of light are filtered out for each pixel using a filter. For example, the red pixel has a filter that absorbs green and blue light, only allowing red light to pass to the sensor.
Other issues may also affect the efficiency of the pixel. For example, the difference in refractive index between the microlens 14 and the overcoat layer 15 will cause some of the incident photons to reflect off the overcoat rather than be transmitted to the photosensitive element. Typically, the difference between the refractive indices of air (n=1.0) and a typical polymer (n=1.5) overcoat 15 is small, resulting in a small reflective loss. Another, larger reflective loss, however, is generated at the boundary between the inter-level dielectric layer 1 (n=1.5) and the substrate 20 (n=4-5). This is due to the larger difference between the refractive indices of the inter-level dielectric layer 1 (n=1.5) and a typical silicon substrate 20 (n=4-5). Another loss may occur due to the incident light hitting the overcoat at too severe an angle and not being transmitted to the photosensitive element. Additionally, real devices have a quantum efficiency less than 100%. That is, even when photons reach the photosensitive element, a finite number of them do not produce a signal.
Another issue that plagues image sensors is crosstalk. Crosstalk is a phenomenon by which a signal transmitted in one pixel or channel of a transmission system creates an undesired effect in another pixel or channel. For optical sensors, there are at least three types of crosstalk: (1) spatial optical crosstalk, (2) spectral crosstalk, and (3) electrical crosstalk. Spatial optical crosstalk occurs when the pixel size approaches the wavelength of visible light. Diffraction causes a sharp increase in the amount of light that reaches adjacent photodiodes rather than the desired photodiode. Spectral crosstalk is when light that should have been blocked by a color filter manages to pass through the filter. Electrical crosstalk is when photo-generated electrons travel to adjacent pixels through the silicon substrate.
The development of nanoscale technology and in particular the ability to produce nanowires has opened up possibilities of designing structures and combining materials in ways not possible in planar technology. One basis for this development is that the material properties of a nanowire makes it possible to overcome the requirement of placing a color filters on each photo diode of an image sensor and to significantly increase the collection of all the light that impinges on the image sensor.
Nanowires of silicon can be grown on silicon without defects. In US 20040075464 by Samuelson et al. a plurality of devices based on nanowire structures are disclosed.